What does Openshift mean when it says it's "Evicted" my pods - and why does it keep doing so?

Question

I am using OpenShift 3 Starter on OpenShift Online to run a small web project. It had been running fine for the last two months and suddenly stopped running yesterday. I found the deployment had simply gotten stuck at "Creating Container..." When I tried to rebuild and restart the project, I ran into this problem. After fixing that, the application now builds.

However, when I try to deploy, the deployment fails. There are no logs. The pod listing says "Evicted."

Why would this be happening?

There is a persistent volume claim in the project, and it is possible that the 1 GB storage space is full. Would this be a reason for eviction? I cannot find any way to check if this is the case or not, as I cannot connect using oc rsh without any running pods, and there seems to be no way to check this fact from the Web Console.

I should mention that I have really struggled to understand the Openshift 3 documentation (which, in contrast to the Openshift 2 documentation, I find very unclear and confusing), so it is possible that I've missed some basic point that is contained in it.

EDIT: Graham Dumpleton in the comments below has pointed out that one of the servers is undergoing an upgrade and another had some scheduled maintenance. However, the maintenance was completed on December 5th (according to the now updated status site). I still cannot find any documentation that says what "evicted" means, so I have no information on how long this problem is going to continue, and what I can do to resolve it.

Dowell

Which OpenShift Online Starter cluster are you on? – Fatuous

Oregon, I think – Dowell

If you are in us-west-1, it has been undergoing an upgrade. Are you on us-west-1 or us-west-2? – Fatuous

It says US West Oregon, no more details than that. But the maintenance on us-west-2 was on the 4th. Would it still be affecting apps? This problem continues till date – Dowell

The URL for the console will tell you whether you are using us-west-1 or us-west-2. The status page at status.starter.openshift.com still shows us-west-1 being in the process of an upgrade and that is causing some out of node memory events to happen, which could well result in evictions. The issue is still being worked on. The notice on us-west-2 back on 4th was notice about upcoming maintenance, but parts of that have actually started. Part of that involves evacuating nodes to upgrade them, which will also cause evictions. – Fatuous

@GrahamDumpleton - thanks. But it's consistently evicting my pods - not randomly, which is what I'd think from your comment - and it's been going on for four days now. If it is an upgrade/maintenance issue, where would I find out how long this is going to continue? It is a production project - albeit a very small one - and this disruption is causing some problems... – Dowell

Yes, it's us-west-2 – Dowell

You might want to refer again to the description of Online Starter. Being a free environment you pay no money for, the Online Starter environments are not meant to be for running production applications. They are for individual learning and experimentation. The limited resources don't make them suitable for production environments. Your application also can be idled (resource hibernation) such that it can't run 24/7. Details of what you get are found at openshift.com/pricing/index.html For a production system, Pro should be used. – Fatuous

Answer 1

And, after four days, it started working again today. Clearly @GrahamDumpleton was right, which is much poorer performance than one would expect from a large company like Red Hat. Hence I'm just leaving this question and answer here in case this happens to someone in future.

Dowell